Note (verbatim, from paper/manuscript_marine.qmd around line 249, under # parameter-dependent \dvv\ and its errors {#sec:results}):
i think this section should systematically quantify bias and variance against the
ground truth for the range of potential parameter choices, or maybe that can be
summarized in a synthesis table.
Context: the note sits right after the section intro (lines 229-246), which lists the components of a well-posed dv/v measurement (estimator family, coherence-based error model, reference window, cross-validation) and states that Section 5's subsections (Estimator family §sec:methods-fig, Aggregating cross-component results §sec:aggregation, and the artifacts/error sections that follow) quantify how each parameter choice moves the recovered dv/v and its error "against a known truth." The ask is to make that quantification systematic across the full parameter space rather than qualitative/scattered across prose, and to consider consolidating it into one synthesis table.
Acceptance criterion: a table (or figure) in paper/manuscript_marine.qmd §sec:results that lists, for each parameter choice covered in the section (estimator family, frequency band/coda window, reference window, substacking, cross-component aggregation/weighting), the induced bias and variance in recovered dv/v relative to the known synthetic ground truth — generated from codameter's existing synthetic framework, not hand-assembled. Closes when that table/figure is in the manuscript and referenced from the section text.
Note (verbatim, from
paper/manuscript_marine.qmdaround line 249, under# parameter-dependent \dvv\ and its errors {#sec:results}):Context: the note sits right after the section intro (lines 229-246), which lists the components of a well-posed dv/v measurement (estimator family, coherence-based error model, reference window, cross-validation) and states that Section 5's subsections (Estimator family §sec:methods-fig, Aggregating cross-component results §sec:aggregation, and the artifacts/error sections that follow) quantify how each parameter choice moves the recovered dv/v and its error "against a known truth." The ask is to make that quantification systematic across the full parameter space rather than qualitative/scattered across prose, and to consider consolidating it into one synthesis table.
Acceptance criterion: a table (or figure) in
paper/manuscript_marine.qmd§sec:results that lists, for each parameter choice covered in the section (estimator family, frequency band/coda window, reference window, substacking, cross-component aggregation/weighting), the induced bias and variance in recovered dv/v relative to the known synthetic ground truth — generated fromcodameter's existing synthetic framework, not hand-assembled. Closes when that table/figure is in the manuscript and referenced from the section text.